Towards Intelligent IoT-NDN Security: AI-Driven PIT Attack Detection and Cache Attack Analysis

Beginning with Named Data Networking (NDN), an early form of information-centric networks, the paradigm of how data is transmitted over a network was changed through the use of ``content-based'' communication instead of ``host-based'', while creating native caching at intermediate points along the path to each destination, and improving upon the security of all previous paradigms. NDN contains many inherent benefits such as caching, security, etc., but like any other paradigm, NDN creates new types of vulnerabilities, particularly within some of the key elements of this paradigm; namely the Content Store (CS), Pending Interest Table (PIT), and the Forwarding Information Base (FIB). Both tables are susceptible to different types of attacks, such as Cache Pollution Attacks (CPA), which decrease cache efficiency, and Interest Flooding Attacks (IFA), which result in network resource depletion due to an excessive number of unfulfilled interest messages being sent across the network. This research will evaluate the effects of CPA and multiple IFA variants, specifically Simple, Switched, and Collusive, via simulation in an IoT-based NDN environment utilizing ndnSIM. Several performance metrics are used to demonstrate the impact of these attacks, such as Cache Hit Ratio (CHR), Interest Satisfaction Ratio (ISR), Timed-Out Interests, Pit Size, and delay. Results demonstrate the ability to provide a unified comparative analysis of CPA and IFA variants (Simple, Switched, and Collusive) within a structured comparative analysis framework. Preliminary results demonstrating the efficacy of Random Forest and Support Vector Machine classifiers for detecting IFAs from simulation-generated datasets demonstrate strong potential for developing future Intelligent Detection Systems, with both classification models achieving an F-score of 0.9923 and 0.9866, respectively.